0:00
This is the fifth module, Moule E in Unit 6, our discussion
about behavior genetic research on general cognitive ability or intelligence.
And in the previous modules we've talked about the evidence that
general cognitive ability is an heritable trait so it is moderately heritable.
And the point of this module is to discuss research aimed at trying to understand,
why is general cognitive ability heritable?
And well, I'm going to touch, actually,
just very briefly on two themes along these lines.
The first is what has been called neurogenetics.
And, I'll get into an illustration of this in a second.
But basically, what neurogenetics involves is trying to identify
the intervening brain systems or structures
that intervene between the DNA we inherit and the behavior we express.
And I'm going to illustrate that approach with general cognitive ability today.
And then the second thing we'll talk about is attempts to try to identify specific
genetic variants that might underlie the heritability of general cognitive ability.
More or less along the same lines as we did with schizophrenia.
But first, let's talk about neurogenetics.
Now this is clearly an oversimplified schematic of what neurogeneticists do.
The notion being that we inherit DNA and somehow the heritability study
suggested that DNA affects individual differences in general cognitive ability.
We presume that it does so by affecting intervening brain phenotypes.
Those could be brain structures or they could be brain processes.
Obviously DNA doesn't directly code for a brain, it codes for, for proteins.
But what neurogeneticists try to do in the case of general cognitive ability, or
indeed in the case of schizophrenia,
is to try to use strategies to identify what those likely brain phenotypes are.
1:56
And there are various criteria you,
that neurogeneticists use in an attempt to identify those important phenotypes.
I'm going to highlight what I think are the most important,
the three most important, which are listed here.
That the, that that brain phenotype is heritable, that not only is it heritable,
but it's also related to the behavioral phenotype we're interested in,
in the case of this module, general cognitive ability.
And finally, that relationship between the brain phenotype and
our behavioral phenotype is due to common genetic effects.
2:31
Now I know this is a very hard slide to read, and
what it, I want you to get is kind of the gestalt here, the overall impression.
What this is is a, is a large-scale meta-analysis published in 2,
2012 of twin studies of various brain measures.
Some of the brain measures, I know you can't read the legend, but
some of the brain measures are things like brain volume, volume for the total brain,
volume for subcortical regions like the hippocampus or the amygdala,
and there are also measures of brain function in the,
in this graph, as well as brain connectivity.
3:06
What they've plotted here is the estimate of heritability from twin studies,
that's red, the estimate of the shared environmental effect, blue,
and the estimate of the non-shared environmental effect, green, for,
it looks like, maybe 60, 70, up to a 100 different brain measures.
And the overall impression that I think you would take away from this is not
unlike what, the impression that we got when we looked at behavioral phenotypes.
That the major sources of individual differences,
the major sources of phenotypic variance, are genetic factors, the red,
and non-shared environmental factors, the green.
And the blue doesn't play that much of a role.
3:44
I want to focus in on one of these though, and that's total brain volume.
Total brain volume in this meta-analysis, highly heritable, about 80% heritable,
very little evidence of a shared environmental effect, and
a small estimate for the non-shared environmental effect.
4:13
Now, before answering that question,
which is answered in this slide, I want to emphasize that total brain volume
is probably not the most optimal brain phenotype to look at.
It's not just how big your brain is,
it's probably certain regions in your brain that are most important.
And, and maybe actually even more important than that,
is how those brain regions communicate with one another.
But what we, I'm going to use, I'm going to illustrate the neurogenetic
approach with total brain volume because it's the, the,
the phenotype on which we have most data at this time.
So even if it's, might not be the optimal phenotype,
it's the one I'm going to use to illustrate the approach.
Another meta-analysis published in 2005 reports that, indeed,
how big your brain is, is correlated with how well you do on an IQ test.
The correlation is not overwhelming, but it's certainly statistically significant,
and something that's worthy of being paid attention to, correlation about 0.3.
The second criterion is met.
The third criterion asks, is this correlation due to common genetic factors,
or is it due to common environmental factors?
If you listened to the supplemental lecture that we had in
Unit 3, then you'll know that answering that last criterion, or
addressing that last criterion, is really
something that people use multivariate behavioral genetic approaches to do.
I'm going to illustrate it here as well.
5:45
In this study,
this is a study published by Danielle Posthuma in Nature Neuroscience in 2002.
Danielle Posthuma is a very prominent neurogeneticist from from the Netherlands.
And in this study, the first thing she reports,
she actually doesn't look at total brain volume so much as how much is gray matter.
Gray matter would include things like your, the neuronal bodies
versus white matter, the tracks between the neuronal bodies, like the axons.
6:11
Those both are correlated with general cognitive ability,
kind of on the order of what we saw on the previous slide.
They're moderate correlations.
That's within individual correlation.
Now the interesting thing about this study is,
she actually did, recomputed the correlation across MZ twins.
So she took one MZ twin's, let's say gray matter volume,
and correlated with his or her co-twin's general cognitive ability.
And what she found is that the correlation between your, the, the amount
of gray matter that you have in your brain, predicts your co-twin's general
cognitive ability as well as it predicts your own general cognitive ability.
And that suggests that the reason that they're correlated
is because of some common genetic effect.
And consistent with that, if we go across DZ twins, then we see that
the correlations are lower, consistent with there being a genetic effect.
7:11
So, the first point here in this module is to illustrate which is of,
of extraordinarily active area of research today.
Trying to go from clinical or
psychological behavioral phenotypes to looking at features of the brain and
at the way the brain functions o try to ident, under,
identify what those underlying neurological systems are that,
that explain the heritability of these behavioral phenotypes.
7:41
The second point I want to talk about then, is general cognitive ability,
IQ is heritable, is maybe 50% heritable in, in, in most studies.
Can we identify what about the DNA sequence underlies that heritable effect?
[SOUND] Now, in this week I've asked you to read this paper.
A paper by Chris Chabris and his colleagues which was published,
I think in 2012.
It's a candidate-gene study of general cognitive ability.
We talked about a major candidate-gene study of schizophrenia in Unit number 5,
the Saunders et al study, and
actually, we spend a lot of time going through that study.
I'm not going to spend a lot of time going through this study
that I also asked you to read.
8:29
The bottom line is the candidate-gene approach, we talked about this last week,
did not really produce much yield in research on schizophrenia.
Its record is pretty similar in general cognitive ability,
and that's really illustrated by this study.
In this study what they did is, they surveyed the literature and
they identified the, and
you can read the paper, but they identified the 10 best candidate-genes.
What researchers thought were the most important candidate-genes for
general cognitive ability.
And they decided to investigate them.
They actually genotyped 12 different DNA variants in these 10 genes.
So two of the genes they, they genotyped two different markers.
They had a reasonably large sample, it was almost 10,000 individuals.
And what they did is, in, on those 10,000 individuals, they had these markers,
or genetic variance genotyped.
They had their measure of general cognitive ability.
They just correlated each across the 12 different genetic markers.
They found no significant associations.
So much like the Saunders study, the candidate-gene approach
has not really yielded much for general cognitive ability.
And indeed it really hasn't yielded much, in my opinion, for
most behavioral and psychiatric phenotypes.
Now that actually raises a question.
Well why hasn't it yielded much?
And I, I'm actually going to re, I'm, I'm going to leave that for
the moment and I'll return to it in the supplemental lecture for this unit.
10:00
So we haven't found much for
general cognitive ability using the candidate-gene approach.
If we recall for schizophrenia they GWAS approach actually began to id,
to identify genetic variance.
Has the GWAS approach similarly been successful with general cognitive ability?
10:18
The first large-scale GWAS of general cognitive ability was published in 2014.
It had about 18,000 individuals pooled over 9 samples,
so this is a meta-analysis.
And those 18,000 individuals were genotyped, or
had information on over 2 million different genetic markers.
So what did the study involve?
Associating each of these 2 million-plus snips
with general cognitive ability in this large, what seems to be a large sample.
Did they find anything?
10:51
Here is the Manhattan plot from that study.
You may recall that, in this case what you're plotting, there's over 2 million p
values plotted here on this minus log to the 10 p value scale.
So the higher the peak, the more significant the results.
But to call something significant it has to be above 7.3, and
in this study nothing actually went above 7.3.
So at this point, for general cognitive ability,
the candidate-gene studies haven't worked and GWAS hasn't worked.
11:23
Maybe, and some critics have claimed this,
maybe the twin studies and adoption studies are wrong,
it's not a heritable trait, and that's why we're not finding anything.
My personal opinion is that 17,000 or 18,000 individuals,
although it seems gigantic, given what we now know about the human genome,
it's probably too small to really identify anything.
If you remember from last week when we talked about schizophrenia,
recall that the first large scale GWAS of schizophrenia found nothing.
But the most recent GWAS of schizophrenia that was much larger than the first one
found over 100 different genetic variants.
So maybe what this is telling you, us, and it's actually my interpretation anyway of
the literature, is that no, we haven't found anything, but we don't have,
even though 18,000 seems extraordinarily large, a large enough sample.
And one reason to, at least in my opinion, to believe that maybe 18,000
isn't a large enough sample, is another GWAS of a related phenotype.
This is a study that was published in 2013 in Science.
It's a large consortium, it involves over 120,000 participants.
And actually there must, there must be I don't know,
maybe 200 authors on the paper including myself.
I'm one of the authors on the paper because we contributed data
to this effort in this massive meta-analysis.
In this case the phenotype for this GWAS is how long you go to school.
And there are actually two phenotypes, whether or not you
attained a college degree and secondly, how many years of education you had.
13:06
And when we analyze the multiple millions of snips with respect to those two types
of phenotypes, here are the Manhattan plots, and this is looking a little bit
more like Manhattan than the previous one for general cognitive ability.
Again recall on this scale, this minus log peak scale,
to be significant it has to be greater than this line, 7.3.
It turns out that there's two significant signals, as it were, for whether or
not you attained a college degree, and one significant signal for how many years of
education you have, chromosome one, chromosome two, and chromosome six.
So actually three variants were found in this.
13:48
But not surprisingly hopefully for
you it's not surprising at this point, the magnitude of the effects
of these variants on the phenotype were extremely small.
For example, that variant that was associated for
how far individuals went in school was a,
the effect was two additional months for each allele you had on that variant.
So it's very, very small effects that we're talking about.
But that's consistent, entirely consistent with what we saw for
schizophrenia, for what we talked about in height, and indeed,
for what human geneticists are finding for one trait after the other.
14:27
Now the last thing before I move on to try and summarize here,
I want to say about the study is, when we looked at these three variants here, and
we correlated them with a measure of general cognitive ability in a smaller
sample, indeed they were significantly correlated with general cognitive ability.
So even though the original GWAS was on educational attainment,
it appears that they're actually identifying variants that are associated
with general cognitive ability.
Giving me, at least people like me, optimism that as samples get larger,
we will, like in the case of schizophrenia, begin to identify those
genetic variants that underlie the heritability of general cognitive ability.
15:07
So, what have we learned about specific genetic factors underlying heritability?
We've talked about two phenotypes in detail, schizophrenia and
now general cognitive ability.
We've found out a couple things.
First of all, and again, I'm going to return to this in the supplemental
lecture, the candidate-gene approach that really dominated
in behavioral genetics and psychiatric genetics for
almost 20 years has not really been particularly successful.
I don't want to par, completely dismiss the candidate-gene approach.
I think it's still a viable approach.
However, during that period of our,
of, of the history of our science, it didn't actually yield much.
Secondly, we are finding results with GWAS.
But when we're finding things with GWAS the affects are very very small,
therefore it takes massive sample sizes to find those effects.
And even with those massive sample sizes, a we, as we saw with height or
schizophrenia, we don't really identify all the relevant genetic variants.
Most of the var, most of the heritability is missing.
And one question that the field is grappling with, and
there's actually quite a bit of debate on this, and we'll come back to it actually
in the last week in this course, is whether or not it's worthwhile to invest
all these resources in identifying variants that have very small effects, and
account for a small portion of the varia, of the variants and the heritability.
The final thing is that, is that we, we've talked about is that behavioral phenotypes
aren't all that different coming out of GWAS as are physical phenotypes.
That is, just like physical phenotypes, the variants for phys,
psychological and psychiatric phenotypes are small and they, it, when we
accumulate all the ones we know about, still most of the variance is missing.
Matt McGueRegents Professor
